Table 4.

Model fit statistics and class assignment diagnostics

Goodness‐of‐fit statistics
Model	Observations	Log likelihood	DF	AIC	BIC	Entropy
1‐Class	947	−11363.3	19	22764.6	22856.8	—
2‐Class	947	−1002.1	37	22078.2	22257.7	0.79
3‐Class	947	−10878.1	55	21866.1	22133.0	0.78
4‐Class	947	−10789.2	73	21724.5	22078.8	0.76

Assignment accuracy diagnostics
Classes	Probability of class membership	Proportion assigned to class	AvePP	OCC
Older high‐risk	0.196	0.196	0.92	47.17
Younger high‐risk	0.241	0.250	0.81	13.43
Younger moderate‐risk	0.364	0.360	0.91	17.67
Older low‐risk	0.199	0.193	0.82	18.34

AIC, Akaike Information Criteria; AvePP, Average (mean) Posterior Probability of Assignment, ≥0.70 indicates high assignment accuracy [42]; BIC, Bayesian Information Criteria, with lower values signifying a better fit [17]; DF, degrees of freedom; OCC, Odds of Correct Classification, OCC > 5 represents high assignment accuracy [42].

The closer the entropy value is to 1, the stronger the separation between classes [43].

The 4‐class model does not meet the conditional independence assumption; however, experts have emphasized that this assumption is more difficult to meet when classifying based on behavioral indicators, and that conditional independence must be balanced with interpretability [44, 45].

We did not calculate the Likelihood‐Ratio test for each model, since this test is based on the chi‐squared statistic which requires observed and expected values and can only be used when all indicators are categorical [41].